Schema drift is a deploy problem, not a data problem
A column changes type at 10:42. By 10:43 three dashboards are wrong. The fix belongs in your deploy, not your warehouse.
Ines Duarte
Founding engineer
6 min
Most schema incidents start the same way. Someone ships a migration on the app side, and the data team finds out from a broken dashboard.
Treat schemas like code
Sluice versions every source schema. A new nullable column propagates on its own. A type change or a dropped column pauses that one table and posts the diff to Slack, so the person who shipped it can approve or roll back.
Nothing is lost while paused
Paused tables keep buffering. When someone presses Resume we backfill the gap in order, and the warehouse never sees a half applied change.